I have a large number of samples which represent Manchester encoded bit streams as audio signals. The frequency at which they are encoded is the primary frequency component when it is high, and there is a consistent amount of white noise in the background.
I have manually decoded these streams, but I was wondering if I could use some sort of machine learning technique to learn the encoding schemes. This would save a great deal of time manually recognizing these schemes. The difficulty is that different signals are encoded differently.
Is it possible to build a model which can learn to decode more than one encoding scheme? How robust would such a model be, and what sort of techniques would I want to employ? Independent Component Analysis (ICA) seems like could be useful for isolating the frequency I care about, but how would I learn the encoding scheme?